The present invention relates to a fabrication method for a flash memory, and more particularly relates to a fabrication method for a self-aligned flash memory.
A computer with advanced function must process a large amount of data, which requires more memory. At the end of 1980, Intel developed the xe2x80x9cflash memoryxe2x80x9d which is based on the same framework as the electrically erasable programmable read only memory, EEPROM. Because of the number of times of data is stored, read and erased is not limited, and thus the flash memory has become the fastest memory developed.
Nonvolatile memory, especially flash memory, is becoming more and more important in many applications. In recent years, the demand for flash memory has exceeded that for other kinds of memory. Therefore, how to develop the new generation of technology and how to reduce the area of flash memory are the challenges facing engineers.
The framework of the flash memory is similar to that of the EPROM, having a stacked gate structure comprising a control gate and a floating gate. The only difference is the thin and high quality tunnel silicon oxide layer under the floating gate. When the program is written into the flash memory, the hot carrier tunnels through the thin tunnel silicon oxide layer to the floating gate, which is similar in operation method to the EPROM. When erasing the program, a negative voltage is applied to the control gate to force the carrier trapped in the floating gate to tunnel through the thin tunnel silicon oxide layer to the source area.
Although flash memory and the EEPROM have the same framework, their erasing methods are different. The volume of the flash memory is much smaller than that of the EEPROM. Moreover, the time required to erase data is about one to two seconds. This is much faster than that of the EPROM, which requires twenty minutes and ultraviolet (UV) rays as the source to erase the data. Therefore, the flash memory owns has advantages over EPROM and EEPROM.
FIG. 1 illustrates a flash memory fabricated by the conventional method. It comprises using a photomask (not shown in the figure) to define the active area 16 on the P-type substrate 10. Next, a dry etching is performed to create a recess in the wafer for formation of shallow trench isolation 17 (STI), the depth thereof being about 200 to 400 nm in the substrate 10. Then the photomask is removed and a stacked gate structure is formed over the substrate 10. The stacked gate structure includes a tunnel silicon oxide layer 12 and a floating gate 13. Next, an ONO (oxide/nitride/oxide) insulating layer 14 and control gate layer 15 are sequentially formed over the tunnel silicon oxide layer 12 and a floating gate 13. It is noted that additional photomask is required to define the area of the floating gate 13. Therefore, at least two different kinds of photomasks are required to form the floating gate 13 and active area 16 in the conventional method.
It is assumed that a V voltage is applied to the control gate layer 15 and a zero voltage is applied to the substrate 10. Because the substrate 10, the floating gate layer 13 and the control gate layer 15 present the capacitance coupling state, the floating gate voltage (VFG) is determined according to the following formula:       V    FG    =            V      xc3x97              C        2                            C        1            +              C        2            
C1 is the capacitance between the substrate 10 and the floating gate 13. C2 is the capacitance between the floating gate 13 and the control gate layer 15.
When the channel length is reduced to 0.18 xcexcm, even to 0.13 xcexcm, the conventional non-using self aligned fabrication technology of flash memory experiences misalignment between the floating gate layer 13 and active area 16. On the other hand, this kind of flash memory that injects a carrier into the floating gate layer to store data has a faster injection velocity if the tunnel silicon oxide layer 12 is thinner or the floating gate layer 13 covers more of the source area. However, the thinner tunnel silicon oxide layer 12 also reduces the data retention time.
On the other hand, under constant applied voltage, the voltage VFG of the floating gate layer 13 can be increased by increasing the capacitance of C2 such that the applied voltage can be coupled more to the floating gate layer 13. That is, the area between the floating gate layer 13 and the control gate layer 15 is increased. The foregoing method can lower the applied voltage to generate the tunnel current of tunnel silicon oxide layer 12. However, because of the limits of the photolithography process, the isolation distance between the floating gates is also limited. Therefore, a greater area of shallow trench isolation 17 is necessary because of the limit of the photolithography process when using the method of increasing the covering area to increase the capacitance of C2. This, in turn, increases the area of the flash memory. A conventional method of using the spacer process to overcome the foregoing drawback also requires a photolithography process so that the whole reduced area of flash memory is still limited.
Therefore, resolution of the misalignment problem between the floating gate 13 and active area 16 and how to increase the capacitance of C2 but not increase the whole area of flash memory is the greatest challenge facing engineers today.
In accordance with the foregoing description, two photo masks are respectively required to form the floating gate layer and the active area of the conventional method. Precise control is very difficult to attain during the alignment process. Therefore, one of the purposes of the present invention is to provide a self-aligned process for flash memory to resolve the misalignment problem.
Another purpose of the present invention is to provide a new kind of floating gate structure and a related process method. The method may increase the coupling area between the floating gate layer and the control gate layer to raise the coupling capacitance thereof but not increase the whole area of the flash memory.
The process method of the present invention comprises these steps. Firstly, a shallow trench isolation is formed on a semiconductor wafer to form the active area islands. Next, the shallow trench isolation is etched to form an indented shallow trench isolation that forms the active area islands. A silicon oxide layer is formed over the indented shallow trench isolation and the surface of the wafer. An anisotropic dry etching step is performed on the silicon oxide layer to form dielectric spacers surrounding the sides of the active area. Then, a thermal process is performed to form a thin tunnel silicon oxide layer. A doped polysilicon layer is deposited as a floating gate over the tunnel silicon oxide layer, dielectric spacers and indented shallow trench isolation. Next, the indented shallow trench isolation between the active area islands is filled with a silicon nitride layer. Then, a thermal oxide layer is grown on the polysilicon layer not covered by the silicon nitride layer. Hot phosphoric acid is used to remove the silicon nitride layer. Using the thermal oxide layer as a mask, the polysilicon layer is etched to isolate the floating gate. Next, a dielectric layer comprising oxide-nitride-oxide (O/N/O) is sequentially deposited on the surface. A doped polysilicon gate is formed as a control gate over the dielectric layer. Then, conventional photolithography and etching processes are performed on the gate structure to form the word line. Finally, a source and drain area ion implant is performed, after which the self-aligned flash cell is complete.
Formation of a floating gate according to the present invention does not require a photolithography step so the method does not suffer misalignment between the floating gate and the active area. On the other hand, the active area islands are formed by etching the shallow trench isolation. Therefore, the coupling area between the floating gate layer and the control gate layer may be increased by raising the etching depth so the coupling capacitance is increased but does not affect the whole area of flash memory.